Revisiting Russell’s Theory of Descriptions By Patrick Henning (California State University, Fullerton)

Abstract: Bertrand Russell’s theory of definite descriptions played a significant role in the development of philosophy of language. However, the shift from semantics to pragmatics in the narrative of language philosophy seemed to leave Russell’s theory in the past as an important but obsolete stepping stone. There is a chance that Russell may have been dismissed too casually, and if so, the grounds on which his theory is rejected must be carefully re-evaluated. In this paper I examine two problems with Russell’s theory that extend beyond the most well-known direct criticisms. In particular, I investigate problems with Russell’s approach to egocentricity as well as his treatment of descriptions of fictional subjects. I then proceed to explore the high-level implications of these problems.

Bertrand Russell’s theory of definite descriptions, first presented in “On Denoting”, played a significant role in the development of philosophy of language. The shift from semantics to pragmatics (initiated in part by P. F. Strawson’s reply to Russell: “On Referring”) in the narrative of language philosophy seemed to leave Russell’s theory in the past as an important stepping stone that has since been superseded by newer theories. “Mr. Strawson on Referring”, Russell’s subsequent attempt to uphold his theory in the face of changing trends, is “seldom taken seriously, at least not as a well-reasoned and successful defense against Strawson's superior criticism” (Austin, 531). However there is a chance that Russell may have been dismissed too casually, and if so, the grounds on which his theory is rejected must be carefully re-evaluated. In this paper I will first examine two problems with Russell’s theory that extend beyond Strawson’s direct criticisms. I will then proceed to explore the implications of these problems. In order to be reasonably charitable throughout the course of criticism, our initial point of engagement with Russell will be within his own domain of logical analysis. Russell’s theory was constructed, in part, as a solution to the problem of empty denotation. There are names and other denoting phrases which seem to refer to nothing, and thus present a puzzle for language philosophers. An example of an empty denoting phrase is ‘the present King of France’, as there is currently no King of France to refer to. Empty denotation is at the root of confusion regarding how to analyze sentences such as the following: (S1): “The present King of France is wise” Since there is presently no King of France, what is the correct way to interpret this sentence? There are several potential answers. According to Alexius Meinong, if a denoting phrase is grammatically correct then there is a guaranteed referent. In the case of no real referent it is an unreal object. Russell is unsatisfied with such an answer due to his belief in the importance of a robust sense of reality which excludes nonexistent objects (“On Denoting” 885). Gottlob Frege put forth the idea that denoting phrases 49

have sense (meaning) as well as denotation (reference). An empty denoting phrase refers to the null class, and has a meaning despite its failure to refer. However Russell felt that a distinction between meaning and denotation is erroneous: if ‘the present King of France’ retains meaning in the absence of a referent, then we are inclined to say that S1 is nonsense (due to its seeming lack of truth or falsity) (“On Denoting” 878). Russell viewed S1 as a meaningful sentence, and by his reasoning, (P1): a sentence is meaningful if and only if it has a truth value. So the demonstration of the meaningfulness of S1 would require a demonstration of either truth or falsity. Russell argued that the ambiguity of sentences in the vein of S1 is due to fact that the structure of ordinary language is at odds with the structure of logical propositions. Contrary to Frege, Russell held that denoting phrases (which are not complete sentences on their own) do not in themselves have a meaning. If Russell was correct, then the correct explicit rendering of a proposition adjusts the role of the denoting phrase, yet preserves the meaning. In order to prevent S1 from being nonsensical, Russell formulated it in the following way: (S1*): “There exists one and only one x such that x is the present King of France and x is wise.” S1* serves Russell’s goal as it is both sensical and false. This mode of interpretation allows Russell to treat language which involves empty denotation as sensical without having to admit Meinongian nonexistent objects into his ontology. An important assumption that the formulation of S1* depends on is the premise that (P2): “all uniquely referring expressions entail a uniquely existential claim” (Austin, 532). Strawson takes issue with P1 (the contingence of meaning upon truth value), which can be challenged by asserting that sentences such as S1 actually have no truth value yet are meaningful nonetheless. Strawson contends that P2, and the special treatment given by Russell to empty denoting phrases, is part of an ad hoc attempt to preserve the integrity of P1 in the face of sentences like S1. Although S1* is logically compatible with P1, for Strawson the fact remains that meaning is not dependent on truth or falsity. Central to Strawson’s account of meaning (and his criticism of Russell) is the distinction between a sentence and the use of a sentence. For Strawson the meaning of a sentence is simply the rules which govern its correct usage. To illustrate the difference between a sentence and its uses we can consider several examples. First, we imagine that S1 is uttered during the reign of Louis XIV, and that Louis XIV is wise. In this case, the assertion made by the use of S1 would be true. Next, we imagine that S1 is uttered during the reign of Louis XV, and that Louis XV is not wise. Here the assertion made via the use of 50

S1 is false. Finally, we imagine that S1 is uttered in the present. Russell would say that in this situation the sentence is false. Strawson, on the other hand, would say that in this case S1 has been used vacuously and is neither true nor false. We can’t ascribe truth or falsity to a sentence or every use of a sentence; only to assertions made in particular correct uses of sentences. At this point it may appear that Strawson is the clear winner in this dialectic. However James Austin’s examination of “Mr. Strawson on Referring” presents a compelling reason why Russell may have been at least partially correct. Russell’s theory is dismissed by Strawson based on P1, but Austin suggests that Russell was not asserting P1 in the first place. Austin argues that Russell is dealing strictly with propositions, not sentences, and that Russell would not assent to P1 as it has been constructed here; he is not making a normative claim about strings of words in all possible forms and contexts. Instead of P1, Russell may have been claiming that (P1*): a sentence is meaningful when and only when it is being used to make a true or false assertion. P1* brings Russell’s theory closer to Strawson’s in terms of how truth values are assigned. Of course it is important to examine precisely how it is that Russell’s theory handles differences in truth values between uses of sentences. The Russellian account of variations in sentence use relies on the concept of egocentric particulars. Egocentric particulars are "words of which the meaning varies with the speaker and his position in time and space" (Austin, 534). The idea is that egocentric particulars allow for variations in the referent of a description. Variable reference accounts for the difference between a sentence in isolation and a sentence in use in a specific context. To show that Strawson’s criticisms are taken care of by egocentric particulars, Austin provides an adjusted version of S1 which removes the egocentricity: (S2): “The king of France in 1905 is wise”. S2 can only be used to assert one thing, thus its meaning has been fixed via the removal of egocentricity. It appears that the claim can no longer be made that Russell’s theory overlooks the difference between sentences and their use. Austin’s defense of Russell is a strong one (for now we will grant its validity) and in its wake a valid criticism of Russell will likely need to somehow demonstrate the weakness of Russell’s use of egocentricity. In “A Strawsonian Objection to Russell's Theory of Descriptions”, Murali Ramachandran presents an interesting criticism of Russell’s theory that goes beyond the territory explicitly covered by Strawson. Ramachandran’s approach circumnavigates a difficulty (related to Austin’s point) faced by the 51

Strawsonian. Strawson would insist that Russell’s theory requires sentences with multiple potential referents to be false (when they could in fact be found true based on contextual clues). The Russellian reply to this argument would be that the context of utterance will provide the missing piece of an incomplete description (regarding an egocentric particular), which seems to place the debate at an impasse. The possibility for Russell’s theory to be embellished in the aforementioned way necessitates an alternative angle of attack. Ramachandran’s goal is to find an example of an utterance that, under Russellian analysis, has a truth value even though it shouldn’t. In search of such a case, we can imagine a situation in which the sentence (S3): “The table is covered with books” is uttered. The context of utterance in this case is a room with many tables (some covered with books, others not), and the speaker has not pointed to any one table in particular. The intended meaning of the statement in this instance is unclear. S3 is certainly meaningful in the Strawsonian sense. However, Ramachandran argues that the utterance of S3 described previously is unintelligible: there is no way for us to determine the truth value of the utterance as we have no idea which table the speaker is referring to. The unintelligibility of this utterance of S3 seems to have been established, so if Russell’s theory cannot account for this, then there is indeed a problem. We can attempt to find a problem by analyzing S3 à la Russell. In Russell’s view, the correct way to formulate S3 is as follows: (S3*): “There is exactly one table and whatever is a table is covered with books.” The problem with S3* is that it is, unlike S3, intelligible; the utterance of S3* would be ostensibly false as there is more than one table. But our intuition tells us that we cannot so easily declare S3 to be false. If there was some beyond-semantic factor of the situation that indicated the speaker’s belief in the presence of only one table, we could comfortably declare the utterance of S3 false. Alternatively, if the speaker had pointed to a particular table with no books on it, we would also have reason to ascribe falsity. However the situation has no such elements (so we cannot accurately determine the speaker’s intention). Here the discord between ordinary language and Russellian formulation is brought back into full view: the meanings of S3 and S3* are fundamentally different. A Russellian rendering of language excludes the possibility of accurately analyzing cases such as the use of S3 as described previously. Ramachandran is careful to point out that the issue here is different than the issue which Strawson demonstrates with S1, “The [present] king of France is wise”. With S1 the conditions for truth or falsity are easily delineated: there must exist presently a king of France, who will either be wise or unwise. S3 is 52

murkier in terms of the speaker’s intention and is thus a different (and bigger) problem for Russell. This appears to be the case, but we must now consider whether the problem raised by Ramachandran can be applied in such a way as to contend with P1*. So here we return to the subject of egocentric particulars. Austin’s demonstration of the role of egocentricity in Russell’s description theory rests on a particular kind of adjustment to S1: fixing the referent of “the King of France” with a unique qualifier such as “in 1905”. If it is possible to do the same for S3, then the dialectic will shift in favor of Russell. However, I shall show that this is impossible. We start with the sentence S3: “The table is covered with books.” In the vein of Austin’s approach, our next step is to insert a qualifier which removes the egocentricity of the “the table”. To this end, we amend our description of the situation seating the utterance of S3. We now imagine a room with tables arranged in such a way that there is a single table in the center of the room, with other tables scattered around the edges of the room. This revision to the situation produces no change to the original result of Ramachandran’s use of it; the issue of referential ambiguity remains. Our revised situation allows us to posit a new sentence which has no referential ambiguity: (S4): “The table in the center of the room is covered with books”. We now have a sentence which constitutes a proposition that would be either true or false depending on whether the table in question is covered in books. The result of this exercise must not be taken for granted. While at first glance Austin’s point seems to have been reinforced, the fact remains that S4 is not equivalent to S3. Furthermore, the distance between meanings of S3 and S4 is even greater than the disparity between S1 and S2. Removing the egocentricity from S3 required more adjustment due to the greater initial ambiguity. The problem remains, yet now in an even harsher spotlight. In continuing to deconstruct Russell’s description theory, we will take a further look at Russell’s treatment of existence claims entailed by P2. One possible criticism is by way of Charles Crittenden, who suggests that formal logic and ontology are separate issues, and that this is not reflected in Russell’s theory. To exhibit Russell’s mistake, the sentence (S5): “the Cyclops lived in a cave” is used. A Russellian judgment would declare this sentence false because there does not exist a Cyclops in the real world, but common sense tells us otherwise. Anyone who has read the Odyssey would be inclined to say that S5 is true, and its truth has nothing to do with the existence of a physically tangible Cyclops. It is for this reason that Crittenden advocates for a syntactical interpretation of the existential quantifier in logical analysis of statements such as S5, as opposed to an existential interpretation. The existential 53

interpretation would require the Cyclops as a scientifically discoverable entity, whereas the syntactical interpretation allows the Cyclops to exist as a fictitious or unreal object. Russell’s endorsement of the existential interpretation is evinced by his stance on Hamlet and Apollo being nonentities (and thus exclusively being the subject of false propositions) as opposed to being unreal (Crittenden, 87). The grouping of fictional and unreal objects together with nonentities leads to a serious discrepancy between the meaning of a sentence in ordinary use and its meaning in formal logic and is therefore best avoided. We now see that there are problems with Russell’s theory of descriptions that are deeper and more subtle than those put forward by Strawson. Not only does his theory incorrectly attribute truth values to certain nonsensical uses of language, but it fails to account for the ability of language to be used to successfully discuss fictional subjects. It is worth noting that Russell is not actually arguing that his logical formulation of statements preserves the exact meaning of those statements in use. Russell wishes to eliminate as much as possible the ambiguity so often present in conventional speech, for the purpose of developing a technical language of philosophy. As such, his formalization of these kinds of sentences can be seen as more of a proposal for a method of refining the way such statements are made than as a method of exact translation. So it may be the case that many have criticized Russell’s theory on the basis of its inability to perform in ways that Russell never intended it to. But ultimately the distance between language as construed by Russell and language in its natural use in the world might be a problem for a theory that is presented as a vehicle for the development of philosophical thought. Strawson rightly pointed out that “ordinary language has no exact logic” (Strawson, 344). However, perhaps the problem lies not with the application of Russell’s method to ordinary language per se, but instead with Russell’s underlying views on what philosophy is and should be. Russell’s line of reasoning is that “physics and chemistry and medicine each require a language which is not that of everyday life”, and that philosophy should thus “make a similar approach towards precision and accuracy” (“Mr. Strawson on Referring” 387). Russell may not have intended the formulation of sentences under his theory to be an exact analogue of ordinary language, but his approach is misguided. Of course, the possibility remains for a method of logical structuring of language to avoid the pitfalls that Russell succumbed to. Such a method or theory will need to account for nuances of language that Russell did not. If a technical language is adopted for philosophical work, the language should be capable of engaging indeterminacy. Rather than deal with vagueness head on, Russell found it “more convenient to define the word ‘false’ so that every significant sentence is either true or false” (“Mr. Strawson on 54

Referring” 388). The ambiguities of natural language which Russell tried to discard are not merely inconveniences to be avoided. In fact, vagueness and implication are key components of many nonwestern philosophical traditions. Graham Priest, in an article titled “Beyond true and false”, investigates alternative systems of logic that incorporate ambiguity. Specifically, Priest examines the potential for a synthesis of plurivalent logic with Buddhist metaphysics. Possible values for a proposition in such a system of logic include True and False, but also Neither true nor false, and Ineffable (Priest). Of particular interest for the present exploration is N. Rather than assigning F to an ambiguous use of S1 (“The present King of France is wise”), we can give it the value N. This is just one possible way in which a logical approach to language can be adequately equipped to deal with meanings beyond those afforded by a rigid True-False dichotomy. A technical language that is not restricted to a truth binary is more useful for philosophical discourse than a Russellian language. For Russell to suggest that a logically precise, specialized technical language based on western notions of truth and falsity should be a standard for philosophy is a mistake for many reasons. Several problems with the rigid dichotomy approach have been presented here, however there are many more. Within the span of this paper, it has been shown that Russell’s theory of descriptions is inadequate for the analysis of ordinary language, and that this inadequacy extends beyond the domain of ordinary language into philosophical discourse as well. It is important that the reasoning behind these conclusions is more charitable to Russell’s view than Strawson was. Russell lost this round, but it was a fair fight.

Works Cited Austin, James. "Russell's Cryptic Response to Strawson." Philosophy and Phenomenological Research 38.4 (1978): 531-37. Print. Crittenden, Charles. “Ontology and the Theory of Descriptions.” Philosophy and Phenomenological Research 31.1 (1970) 85-96. Print. Priest, Graham. “Beyond true and false.” Aeon Magazine, 5 May 2014. Web. Ramachandran, Murali. "A Strawsonian Objection to Russell's Theory of Descriptions." Analysis (1993): 209-12. Print. Russell, Bertrand. “Mr. Strawson on Referring.” Mind 66.263 (1957): 385-389. Print. Russell, Bertrand. “On Denoting.” Mind 114.456 (2005): 873-887. Print. Strawson, P. F. "On Referring." Mind 59.235 (1950): 320-44. Print.

55